U.S. patent number 7,649,891 [Application Number 11/295,481] was granted by the patent office on 2010-01-19 for combined customer-flow and quality-class-based scheduling method and apparatus for scheduling transmission-link capacity between packet-switched traffic flows.
This patent grant is currently assigned to Tellabs Oy. Invention is credited to Marko Kalevi Kulmala, Janne Vaananen.
United States Patent |
7,649,891 |
Vaananen , et al. |
January 19, 2010 |
Combined customer-flow and quality-class-based scheduling method
and apparatus for scheduling transmission-link capacity between
packet-switched traffic flows
Abstract
The invention relates to a method and apparatus for scheduling
transmission-link capacity between packet-switched
telecommunications flows, in such a way that it is possible to
guarantee that packets representing delay-critical traffic will be
scheduled to the transmission link (S0) before packets representing
non-delay-critical traffic and, in addition, that it is possible to
limit the joint transmission speed of specific service-quality
classes of an individual customer flow. The invention is based on
the service-quality classes being separated into two different
categories: a service-quality-class based scheduling branch (C) and
a customer-flow based scheduling branch (F). When selecting the
next packet to be forwarded, one of the two said branches (C or F)
is selected and in the service-quality-class based scheduling
branch a service-quality class (CoS) is selected and, within the
selected service-quality class, a customer flow (CF) is selected
and in the customer-flow based scheduling branch a customer flow
(CF) is selected and, within the selected customer flow, a
service-quality class (CoS) is selected.
Inventors: |
Vaananen; Janne (Espoo,
FI), Kulmala; Marko Kalevi (Espoo, FI) |
Assignee: |
Tellabs Oy (Espoo,
FI)
|
Family
ID: |
33547958 |
Appl.
No.: |
11/295,481 |
Filed: |
December 7, 2005 |
Prior Publication Data
|
|
|
|
Document
Identifier |
Publication Date |
|
US 20060126637 A1 |
Jun 15, 2006 |
|
Foreign Application Priority Data
|
|
|
|
|
Dec 9, 2004 [FI] |
|
|
20041586 U |
|
Current U.S.
Class: |
370/395.21;
370/428; 370/412; 370/395.43 |
Current CPC
Class: |
H04L
47/2441 (20130101); H04L 47/56 (20130101); H04L
47/2408 (20130101); H04L 47/2425 (20130101); H04L
47/50 (20130101); H04L 47/6215 (20130101) |
Current International
Class: |
H04L
12/28 (20060101); H04L 12/54 (20060101) |
Field of
Search: |
;370/351,389,395.1,396,397,398,395.21,395.3,395.41,395.42,395.43,395.5,395.51,395.52,428,429,412,468,395.2,395.4,352 |
References Cited
[Referenced By]
U.S. Patent Documents
Foreign Patent Documents
|
|
|
|
|
|
|
1 093 266 |
|
Apr 2001 |
|
EP |
|
1093266 |
|
Apr 2001 |
|
EP |
|
WO-01/24428 |
|
Apr 2001 |
|
WO |
|
WO-02/33428 |
|
Apr 2002 |
|
WO |
|
WO-03/039052 |
|
May 2003 |
|
WO |
|
Other References
RFC 2475, Internet Engineering Task Force (IETF): An Architecture
for Differentiated Services, Dec. 1998. cited by other .
Pawan Goyal, Harric M. Vin, Halchen Chen. Start-time Fair Queuing:
A Scheduling Algorithm for Integrated Services Packet Switching
Networks. Technical Report TR-96-02, Department of Computer
Sciences, University of Texas Austin, 1996. cited by other .
P.F. Chimento, Standard Token Bucket Terminology,
<http://qbone.internet2.edu/bb/Traffic.pdf>, May 18, 2000.
cited by other.
|
Primary Examiner: Nguyen; Hanh
Attorney, Agent or Firm: Birch, Stewart, Kolasch &
Birch, LLP
Claims
The invention claimed is:
1. Method for scheduling transmission-link capacity between
packet-switch telecommunications flows in a telecommunications
network element, the method comprising: transmitting digital data
as constant or variable-length packets, attaching first identifier
data to each packet, and classifying said packet, with the aid of
said first identifier data, as belonging to one of at least two
service-quality classes, attaching second identifier data to each
packet, and classifying said packed, with the aid of said second
identifier data, as belonging to one of at least two customer
flows, wherein said second identifier data is defined on the basis
of data inside the packet and/or transported in connection with it
and/or on the basis on the input port of the network element,
associating each service-quality class with either a branch
scheduling on the basis of service-quality class, or to a branch
scheduling on the basis of customer flow, relating a storage buffer
to each customer-flow/service-quality class pair, in which the
packets corresponding to the said customer-flow/service-quality
class pair arriving in the network element are stored, selecting a
packet to be forwarded from among the packets that are offered for
forwarding, and forwarding said selected packet from the network
element over an outgoing telecommunications link, and, when the
transmission has terminated, selecting the following packet,
wherein, when selecting the packet to be forwarded: either the
service-quality-class based scheduling branch or the customer-flow
based scheduling branch is selected and in the
service-quality-class based scheduling branch a service-quality
class is selected and within the selected service-quality class a
customer flow is selected and in the customer-flow based scheduling
branch a customer flow is selected and within the selected customer
flow a service-quality class is selected and the selection between
the service-quality-class based scheduling branch and the
customer-flow based scheduling branch is performed on the basis of
priority, in such a way that the service-quality-class based
scheduling branch is always selected, if there is a packet in even
one storage buffer relating to the service-quality-class based
scheduling branch, otherwise the customer-flow based scheduling
branch is selected.
2. Method according to claim 1, wherein the customer flow is
selected using a weighting-factor based scheduling method (e.g.,
Start Time Fair Queuing, [2]).
3. Method according to claim 1, wherein the service-quality classes
representing delay-critical traffic belong to the
service-quality-class based scheduling branch and the
service-quality classes representing non-delay-critical traffic
belong to the customer-flow-based scheduling branch.
4. Apparatus for scheduling transmission-link capacity between
packet-switched telecommunications flows in a telecommunications
network element, in which the apparatus comprises a reception unit
that receives constant or variable-length packets carrying digital
information, a service-quality class information definition unit
that defines-service-quality class information for the packet, a
customer-flow information definition unit that defines
customer-flow information for the packet, a classification unit
that classifies each service-quality class as belonging to either
the service-quality-class based scheduling branch or the
customer-flow based scheduling branch, a storage buffer for each
service-quality class/customer-flow pair, a selection unit that
selects one packet from among the packets that are offered for
forwarding, a forwarding unit that forwards the selected packet
over the outgoing telecommunications link, and a scheduling branch
selection unit that selects, when the selection unit is selecting a
packet to be forwarded, the service-quality-class based scheduling
branch always, if there is a packet in even one storage buffer
relating to the service-quality-class scheduling branch, and in
other cases the customer-flow based scheduling branch, and that
selects, in the service-quality-class based scheduling branch, a
service-quality class and within the selected service-quality
class, selects a customer flow, and that selects, in the
customer-flow based scheduling branch, a customer flow and within
the selected customer flow selects a service-quality class.
5. Apparatus according to claim 4, wherein the scheduling branch
selection unit includes a weight-factor based scheduler that
selects the customer flow using a weighting-factor based scheduling
method.
6. Apparatus according to claim 4, wherein the apparatus comprises
classification unit that classifies the service-quality classes
representing delay-critical traffic to the service-quality-class
based scheduling branch and classifies the service-quality classes
representing non-delay-critical traffic to the customer-flow based
scheduling branch.
Description
The present invention relates to a method, according to the
preamble of Claim 1, for scheduling transmission-link capacity
between packet-switched traffic flows.
The present invention also relates to an apparatus, according to
the preamble of Claim 4, for scheduling transmission-link capacity
between packet-switched traffic flows.
In the present document, the following abbreviations are used in
the descriptions of both the prior art and the invention: CF
Customer Flow, a traffic flow, which represents, for example, a
specific virtual network (VPN, Virtual Private Network), CoS Class
of Service, DSCP Differentiated Service Code Point, the information
carried by a packet concerning the class of service to which the
packet in question belongs, PCQ Per Class Queuing, queuing and
scheduling discipline, in which queuing is implemented
service-class-specifically and transmission-capacity scheduling is
based on service-quality classes, PFQ Per Flow Queuing, queuing
discipline, in which queuing is implemented customer-flow and
service-class-specifically, PFQ-CS Per Flow Queuing--Class-based
Scheduling, queuing and scheduling discipline, in which queuing is
implemented customer-flow and service-class specifically and
scheduling is implemented service-class specifically, PFQ-FS Per
Flow Queuing--Flow-based Scheduling, queuing and scheduling
discipline, in which queuing is implemented customer-flow and
service-class specifically and scheduling is implemented
customer-flow specifically.
In a packet-switched telecommunications system, it is often
advantageous for the telecommunications packet (Pkt, FIG. 1,
hereinafter `packet`) being transmitted to be classified as
belonging to different classes of service (CoS) according to the
kind of requirements that the applications using telecommunications
service have, and, on the other hand, according to what kind of
service-quality agreements the telecommunications service provider
has made with its customers. For example, in the case of a
conventional telephone service, it is important for the
data-transmission speed of the application has sufficient time
available and that the transmission delay is sufficiently small,
and that the variation in the delay is sufficiently small. In a
telephone application, there is no benefit in being able to
momentarily increase the data-transmission speed provided for the
application, if the loading of the telecommunications network is
low at the time in question. On the other hand, when downloading
web-pages, for example, it is very advantageous if the available
capacity of the network can be exploited in full even
momentarily.
By way of example, a situation will be examined, in which a
telecommunications service provides the following kinds of
service-quality classes: Class of service CoS3: the highest
priority class for applications, for which the data-transmission
delay and the variation in the transmission delay are sought to be
minimized, but the momentary data-transmission speed provided for
the application is not increased, even though the loading of the
telecommunications system may be low at the time in question. The
amount of traffic representing the service-quality class CoS3 is
assumed to be limited. Class of service Cos2: the second-highest
priority class for applications, for which the data-transmission
delay and the variation in the transmission delay are sought to be
minimized, but the momentary data-transmission speed provided for
the application is not increased, even though the loading of the
telecommunications system may be low at the time in question. The
amount of traffic representing the service-quality class CoS2 is
assumed to be limited. CoS1: for applications, for which
upper-limit values are not guaranteed for the transmission delay
and variation in the delay, but for which a specific minimum
transmission capacity is guaranteed and for which the available
capacity of the data transmission system at each moment is
exploited. CoS0: for applications, for which upper-limit values are
not guaranteed for the transmission delay and variation in the
delay, and for which a data transmission capacity is not
guaranteed, but for which the available capacity of the data
transmission system at each moment is exploited.
By way of example, the following situation shown in FIG. 1 will be
examined, in which packets addressed to customers A1 and A2 are
received in the network element NE1 over the data-transmission link
T. The flow of packets addressed to a specific customer forms a
customer flow CF going to the customer in question. The customer
flows CF1 and CF2 going to the customers A1 and A2 are transmitted
from the network element NE1 to the network element NE2 over the
transmission link S0. The customer flow CF1 is forwarded from the
network element NE2 over the transmission link S1 and the customer
flow CF2 over the transmission link S2. Both of the customer flows
CF1 and CD2 contain service-quality classes Cos3, CoS2, Cos1, and
CoS0. The network element NE1 is able to taken the service-quality
class into account when scheduling the packets to the transmission
link S0 (implements the Differentiated Service method [1]). The
network element NE2 processes all the services classes in the same
way (does not implement the Differentiated Service method [1]). The
traffic representing the service-quality class CoS3 of customer A1
is assumed to be limited earlier in the network, before the network
element NE1, to the value MaxA1CoS3 [bit/s]. Correspondingly, the
traffic representing the service-quality class CoS2 is assumed to
be limited to the value MaxA1CoS2. Correspondingly, the upper-limit
values for customer A2 are: MaxA2CoS3 and MaxA2CoS2.
The objective is the following: The capacity of the transmission
link S0 should be divided in the desired ratio between the customer
flows CF1 and Cf2. A packet representing a higher priority will be
transmitted over the transmission link S0 before a packet
representing a lower priority (CoS3 before CoS2, CoS1, and CoS0,
correspondingly CoS2 before CoS1 and CoS0). The traffic shares
representing service-quality classes CoS1 and CoS0 of customer
flows CF1 and CF2, which have not been limited earlier in the
network, must be able to be limited in the network element NE1, so
as not to congest the output ports of the network element NE1 to
the transmission links S1 and S2. This is important, because, in a
congestion situation, when dropping packets NE2 will not be able to
distinguish between packets representing different service-quality
classes.
FIG. 2 shows one way according to the prior art of scheduling the
capacity of the transmission link in a situation like that in the
above example. The procedure shown by FIG. 2 will be referred to in
the following as customer-flow-based scheduling (Per Flow
Queuing--Flow-directed Scheduling, PFQ-FS). The operation of the
system shown in FIG. 2 is as follows: the customer flow (CF) and
service-quality class (CoS) to which a packet arriving in the
system belongs, determines the storage buffer in which the packet
in question is placed. The customer flow can be identified on the
basis of data inside the packet and/or transported in connection
with it and/or of the input port of the network element (NE1). The
service-quality class can be identified on the basis of information
attached to the packet (for example, DSCP=Differentiated Services
Code Point [1]). In the selection process, the customer flow (CF1
or CF2) is selected and, within the selected customer flow, the
service-quality class (CoS3, CoS2, CoS1, or CoS0) is selected. A
customer flow will be a candidate in the selection, if there is a
packet in even one storage buffer relating to the customer flow in
question. The customer-flow selection `select CF` is implemented
using a weighting-factor-based scheduling method (for example,
Start Time Fair Queuing [2]) while the selection of the
service-quality class `select CoS` can be implemented, for example,
as follows: 1) if a packet representing service-quality class CoS3
is offered, always select it, if not, then 2) if a packet
representing service-quality class CoS2 is offered, select it, if
not, then 3) perform the selection between the service-quality
classes CoS1 and CoS0, using a weighting-factor-based scheduling
method (for example, Start Time Fair Queuing [2]). The selection of
the customer flow during a specific selection cycle is prohibited
using a limiter R, if the outgoing traffic representing the
customer flow in question exceeds the speed limits [bit/s or
packet/s] set for it. Speed monitoring can be implemented, for
example, using the traditional Token-Bucket method [3].
The problem with the method shown in FIG. 2 is that, in the
customer flow selection `select CF`, the weighting-factor-based
scheduling method used processes the customer flows (CF1 and CF2)
offered to the scheduler in such a way that both receive the share
of the capacity of the transmission link S0 reserved for them. If,
for example, only packets representing service-quality class CoS3
are going to customer A1 and only packets representing
service-quality class CoS0 are going to customer A2, there is no
guarantee that a CoS3 packet will be scheduled to the transmission
link before a CoS0 packet. Taken generally, the scheduling order is
random. Because the CoS3 packets represent delay-critical traffic,
they should, however, always be scheduled to the transmission link
S0 before the CoS0 packets.
FIG. 3 shows a second way according to the prior art to schedule
the capacity of a common transmission link in the case of a
situation like that in the example described above. Hereinafter,
the procedure shown in FIG. 3 will be referred to as
service-quality-class-based scheduling (Per Flow
Queuing-Class--directed Scheduling, PFQ-CS). The operation of the
system shown in FIG. 3 is as follows: the customer flow (CF) and
service-quality class (CoS) to which a packet arriving in the
system belongs, determines in which storage buffer the packet in
question will be placed. In the selection process, the
service-quality class (CoS3, CoS2, CoS1, or CoS0) is selected and
the customer flow (CF1 or CF2) is selected inside the selected
service-quality class. A service-quality class will be a candidate
in the selection, if there is a packet in even one storage buffer
relating to the service-quality class in question. A customer flow
will be a candidate in the selection, if there is a packet in the
selected service-quality class and a storage buffer representing
the customer flow in question. The selection of the customer flow
`select Cf` is implemented using a weighting-factor-based
scheduling method (for example, Start Time Fair Queuing [2]) while
the selection of the service-quality class `select CoS` can be
implemented, for example, as follows: 1) if a packet representing
the service-quality class CoS3 is offered, it will always be
selected, if not, 2) then if a packet representing the
service-quality class CoS2 is offered, it will be selected, if not,
3) then the selection is made between the service-quality classes
CoS1 and CoS2, using a weighting-factor-based scheduling method
(for example, Start Time Fair Queuing [2]). The selection of the
customer flow during a specific selection cycle is prohibited using
a limiter R, if the outgoing traffic representing the
service-quality class and customer flow being examined exceeds the
speed limits [bit/s or packet/s] set for it. Speed monitoring can
be implemented, for example, using the traditional Token Bucket
method [3].
A problem with the method shown in FIG. 3 is that the customer flow
cannot be limited in such a way that the speed-limited share of the
customer flow would include several service-quality classes and
that the speed limit does not say anything about how the
speed-limited traffic flow is distributed between the various
service-quality classes. In a situation like that of the example
given, it would be advantageous, if it would be possible to
speed-limit the share of the totality of a customer flow (CF1, CF2)
represented by the service-quality classes CoS1 and CoS0, without
separately limiting the shares represented by the service-quality
classes CoS1 and CoS0.
The invention is intended to eliminate the defects of the state of
the art disclosed above and for this purpose create an entirely new
type of method and apparatus for scheduling transmission-link
capacity between packet-switched telecommunications traffic flows.
The invention relates to a method, using which a scheduling
mechanism can be implemented, in such a way that it is possible to
guarantee that packets representing delay-critical traffic will be
scheduled to the transmission link before packets of
non-delay-critical traffic and, in addition, that it is possible to
speed limit the general effect of specific service-quality classes
of a customer flow (such as CoS1 and CoS0 in the example situation
depicted above).
The invention is based on performing the scheduling for specific
service-quality classes (for example, delay-critical
service-quality classes) using quality-class-based scheduling
(PPQ-CS) and, on the other hand, performing scheduling for other
service-quality classes (for example, non-delay-critical
service-quality classes) using customer-flow-based scheduling
(PPQ-FS).
In the system according to the invention, the selection processes
can be implemented using either a programmable processor or
microcircuit mechanisms designed for the operations in question
(Application Specific Integrated Circuit `ASIC`
implementation).
The method according to the invention is characterized by what is
stated in the characterizing portion of Claim 1.
The apparatus according to the invention is, in turn, characterized
by what is stated in the characterizing portion of Claim 4.
With the use of the invention, the advantage is gained over the
solutions according to the prior art that it is possible to
implement the scheduling mechanism in such a way that it is
possible to guarantee that packets representing delay-critical
traffic will be scheduled to the transmission link before packets
representing non-delay-critical traffic and, in addition, it is
possible to speed limit the joint transmission speed of specific
service-quality classes of a customer flow.
In the following, the invention is examined in great detail with
the aid of examples according to the accompanying drawings.
FIG. 1 shows a packet-switched telecommunications network used as
an example for examination. The telecommunications packets
addressed to the customers A1 and A2 are scheduled in a network
element NE1 to a packet forwarding data transmission link S0.
FIG. 2 shows a block diagram of one system according to the prior
art for scheduling the capacity of the transmission link S0 for
traffic flows formed of packets representing the customers A1 and
A2 and the aforementioned service-quality classes (CoS3, CoD2,
CoS1, and CoS0).
FIG. 3 shows a block diagram of a second system according to the
prior art for scheduling the capacity of the transmission link S0
for traffic flows formed of packets representing the customers A1
and A2 and the aforementioned service-quality classes (CoS3, CoD2,
CoS1, and CoS0).
FIG. 4 shows a block diagram of a second system according to the
invention, in an example of an operating environment, prior art for
scheduling the capacity of the transmission link S0 for traffic
flows formed of packets representing the customers A1 and A2 and
the aforementioned service-quality classes (CoS3, CoD2, CoS1, and
CoS0).
The basis of the method according to the invention will become
apparent from the following examination.
In the method according to the invention, the service-quality class
belongs to either to a branch (C, FIG. 4) based on service-quality
class, or to a branch (F, FIG. 4) based on customer flow. In the
service-quality-class-based scheduling branch (C) the scheduling is
implemented by selecting the service-quality class and selecting
the customer flow within the selected service-quality class. In the
customer-flow-based scheduling branch (F), scheduling is
implemented by selecting a customer flow and selecting the
service-quality class within the selected customer flow.
One preferred embodiment of the present invention is a system, in
which the service-quality classes (CoS3 and CoS2) representing
traffic that is delay-critical and assumed to be preliminary
limited are placed in the branch (C) that schedules on the basis of
service-quality class and the service-quality classes (CoS1 and
CoS0) representing traffic that is non-delay-critical and assumed
to be preliminary unlimited is placed in the branch (F) that
schedules on the basis of customer. The selection between the said
branches is performed on the basis on priority, in such a way that
the service-quality-class-based scheduling branch (C) is always
selected, if there is packet in even one storage buffer relating to
the branch (C) in question, otherwise the customer-flow-based
scheduling branch (F) is selected.
REFERENCES
[1] RFC 2475, Internet Engineering Task Force (IETF): An
Architecture for Differentiated Services. [2] Pawan Goyal, Harric
M. Vin, Haichen Chen. Start-time Fair Queuing: A Scheduling
Algorithm for Integrated Services Packet Switching Networks.
Technical Report TR-96-02, Department of Computer Sciences,
University of Texas Austin. [3] P. F. Climento, Standard Token
Bucket Terminology.
http://qbone.internet2.edu/bb/Traffic.pdf2000.
* * * * *
References